Detection of Sub-Design Rule Physical Defects Using E-Beam Inspection

Patterson, Oliver D.; Lee, Julie; Salvador, Dave M.; Lei, Shuen-Cheng Chris; Tang, Xiaohu

doi:10.1109/tsm.2013.2283293

Cited by 12 publications

(2 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Patterning errors ("defects") in semiconductor manufacturing affect the production yield as such imperfections may render computer chips electrically inoperable. Defect inspection is a crucial step in microelectronics fabrication performed using various measurement techniques including optical imaging, 1 optical scattering, 2 scanning electron microscopy, 3 and voltage contrast imaging. 4 Defect inspection is notably different from many other types of measurements in semiconductor manufacturing, separating measurement data into classifications for process control as opposed to translating measurements into physical quantities such as line width or line height.…”

Section: Introductionmentioning

confidence: 99%

Addressing misclassification costs in machine learning through asymmetric loss functions

Barnes

Henn

2023

Metrology, Inspection, and Process Control XXXVII

View full text Add to dashboard Cite

Patterning defect metrology requires data interpretation with classification, each well-suited to machine learning (ML). Defect classification however has notable misclassification costs; mislabeling a defect as nominal has greater impact than the converse. Though quantified costs are not publicly available, total economic misclassification cost (total cost) is optimized across orders-of-magnitude variation in cost ratio C and classification threshold 0.01 <τ< 0.99. Convolutional neural networks are trained using the intrinsically weighted and scaled asymmetric focal losses (AFL, sAFL) with hyperparameter γ with weighted and unweighted binary cross-entropy (wBCE, BCE) functions trained for comparisons. Optimal functions and conditions are identified for reducing total cost. For reproducibility, publicly available ML data sets are surrogates for industrial imaging data. For these data the sAFL mimimizes total cost at τ = 0.5, C ≥ 16. The AFL reduces total cost at 0.1 ≤ τ< 0.5, C > 128. Asymmetric loss functions lower total cost versus wBCE by 15 % to 40 % for 0.2 <τ< 0.5, C< 64. Total economic misclassification cost can be tailored using asymmetric focal losses. Estimations are presented to allow the extension of reported trends to industrial applications with strong class imbalances between defect-indicative and nominal-indicative data.

show abstract

Section: Introductionmentioning

confidence: 99%

Addressing misclassification costs in machine learning through asymmetric loss functions

Barnes

Henn

2023

Metrology, Inspection, and Process Control XXXVII

View full text Add to dashboard Cite

show abstract

“…Many deadly defects are formed during wafer fabrication, such as particles, scratches, pits, and residues, which require high-throughput and high-precision inspection methods to improve chip yields. Two widely used inspection techniques are optical and electron beam 2 , and the latter one has a higher detection sensitivity 3 , however, the main limitation is its slow measurement speed and high cost. In contrast, optical measurement offers non-destructive and high-throughput properties, and is widely used in IC manufacturing process control and yield management 4 .…”

Section: Introductionmentioning

confidence: 99%

Spot-scanning laser scattering system for defects detection of wafer surface

Zhou

Luo

Xiong

et al. 2022

Thirteenth International Conference on Information Optics and Photonics (CIOP 2022)

View full text Add to dashboard Cite

As semiconductors' critical dimension decreases, higher precision inspection instruments are needed to detect defects in the manufacturing process. Optical inspection methods based on light and dark field microscopy can detect defects on large wafer areas well without damaging the wafer, but the minimum detectable defect size is limited because the defect scattering signal is easily buried by the scattering background noise from the wafer's rough surface. To detect submicron defects on wafer surfaces, a spot-scanning laser scattering scheme is developed based on the dark-field scattering technique. Using the Finite Difference Time Domain (FDTD) method and the inspection scheme, an electromagnetic scattering model of the defect on the wafer surface is established, and the defect characteristics and electromagnetic field distribution are simulated. Moreover, the effects of the collecting aperture angle on the signal intensity of defects and the discrimination of defects of different sizes, as well as the effects of the incident angle on the scattered signal intensity of submicron defects, are examined. A spot-scanning laser scattering experimental platform was built, and 200 nm, 1 μm, and 5 μm diameter polystyrene latex (PSL) spheres were deposited on the wafer surface to verify the validity of the proposed method. Signals of the three sizes of spheres were detected in the stitched images with good discrimination of signal intensity, and the signal of the 200 nm PSL sphere displayed a peak signal-to-noise ratio of 32.07 dB. This method provides a reference for further industrialized defect detection systems on wafer surfaces.

show abstract